forked from sweiss93/LengthNestPro
-
Notifications
You must be signed in to change notification settings - Fork 0
/
nest_calculation.py
1351 lines (1113 loc) · 72.2 KB
/
nest_calculation.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
import numpy as np
import time
import math
from PyQt5.QtCore import pyqtSignal, QObject
# from matplotlib import pyplot as plt
# import matplotlib
from column_sort import ColumnSorter
import copy
# matplotlib.use('Agg')
# Create calculate thread class
class CalculateThread(QObject):
def __init__(self, window):
super().__init__()
self.calculation_was_canceled = 0
self.window = window
results_signal = pyqtSignal(list)
@staticmethod
def union(a, b):
for item in a:
if item not in b:
b.append(item)
return b
def find_adj_matrix(self, patterns):
adj_matrix_ = np.zeros((self.num_parts, self.num_parts))
for row_index, row in enumerate(patterns):
for term_index, term in enumerate(row):
if term:
for item_index, item in enumerate(patterns.T[term_index]):
if item:
adj_matrix_[row_index][item_index] = 1
for i in range(self.num_parts):
adj_matrix_[i][i] = 0
return adj_matrix_
# noinspection PyUnusedLocal
@staticmethod
def find_cycling(original_list):
# Reverse the order of the list to make it easier to work with since repetitions would happen at end
reversed_list = original_list[::-1]
# Create generator to find next item in list
index_generator = (i for i, e in enumerate(reversed_list) if e == reversed_list[0])
# Check for first index before entering loop to prevent false positive
first_index = next(index_generator)
# Check for periodic nature in loop until function exits with return
while 1:
# Find next element that matches the first element
try:
next_match = next(index_generator)
except StopIteration:
return False
if next_match < -1:
pass
else:
if np.all(reversed_list[0:next_match] == reversed_list[next_match:(2 * next_match)]) and \
next_match > 10:
return True
def adjust_max_iterations(self, adjustment_factor):
self.max_iterations *= adjustment_factor
if adjustment_factor >= 1:
self.max_iterations += 1
elif adjustment_factor <= 1:
self.max_iterations -= 1
def check_for_bonus_condition(self):
bonus = 1
for check_index in range(len(self.bonus_sublist_sorted)):
if self.bonus_sublist_sorted[check_index] == 1: # Indicates parts that fulfill uff parts
if self.row_copy[check_index + 1]:
continue
else:
bonus = 0
break
elif self.bonus_sublist_sorted[check_index] == -1: # Prevents creation of more uff parts
if not self.row_copy[check_index + 1]:
continue
else:
bonus = 0
break
return bonus
def branch_bound(self, bandwidth, max_parts_per_nest, part_quantities, parts_sublist, mode, current_part_index=0):
# Find pi vector from patterns (Step #2)
if mode == 0:
self.patterns_inv = np.linalg.inv(self.patterns)
self.patterns_trans_inv = np.transpose(self.patterns_inv)
self.ones_vector = np.ones((len(self.patterns[0]), 1))
# elif mode == 1:
# self.patterns_inv = np.linalg.solve(self.patterns.T.dot(self.patterns), self.patterns.T)
# self.patterns_trans_inv = np.transpose(self.patterns_inv)
# self.ones_vector = np.ones((len(self.patterns[0]), 1))
if mode == 0:
self.pi = np.dot(self.patterns_trans_inv, self.ones_vector)
elif mode == 1 or mode == 2:
self.pi = self.nested_lengths / self.nestable_length
# Adjust pi slightly to prioritize longer parts
for i in range(len(self.pi)):
self.pi[i] = self.pi[i] + 0.0001 * self.nested_lengths[i] / self.nestable_length
# Calculate allocations of each pattern
self.allocation = np.dot(self.patterns_inv, part_quantities)
# Find value vector that can be used to prioritize the "usefulness" of nesting each part
self.values = np.divide(self.pi, self.nested_lengths)
if mode == 1:
part_quantities[self.part_index] = self.remaining_part_quantities[self.part_index].copy()
for part in self.unavailable_parts[1:]:
part_quantities[part] = self.remaining_part_quantities[part].copy()
self.values_sublist = self.values[parts_sublist[self.container_counter].astype(int)]
self.pi_sublist = self.pi[parts_sublist[self.container_counter].astype(int)]
self.nested_lengths_sublist = self.nested_lengths[parts_sublist[self.container_counter].astype(int)]
self.part_quantities_sublist = part_quantities[parts_sublist[self.container_counter].astype(int)]
# Sort from "most valuable to nest" to "least valuable to nest" so that optimum solution is reached sooner.
self.index_order = (np.argsort((self.values * -1).transpose()))[0]
self.pi_sorted = self.pi[self.index_order]
self.values_sorted = self.values[self.index_order]
self.nested_lengths_sorted = self.nested_lengths[self.index_order]
self.part_quantities_sorted = part_quantities[self.index_order]
# Sort sublist data from "most valuable to nest" to "least valuable to nest" so that optimum solution is
# reached sooner.
self.sublist_index_order = (np.argsort((self.values_sublist * -1).transpose()))[0]
if mode == 1:
self.sublist_index_order = np.delete(self.sublist_index_order,
np.where(self.sublist_index_order == current_part_index)[0])
self.sublist_index_order = np.insert(self.sublist_index_order, 0, current_part_index)
self.pi_sublist_sorted = self.pi_sublist[self.sublist_index_order]
self.values_sublist_sorted = self.values_sublist[self.sublist_index_order]
self.nested_lengths_sublist_sorted = self.nested_lengths_sublist[self.sublist_index_order]
self.part_quantities_sublist_sorted = self.part_quantities_sublist[self.sublist_index_order]
self.parts_sublist_sorted = np.zeros((self.num_parts - bandwidth + 1, bandwidth))
if mode == 1 or mode == 2:
self.parts_sublist_sorted = np.zeros((1, bandwidth))
if mode == 1:
self.bonus_sublist_sorted = self.bonus_sublist[0][self.sublist_index_order]
self.parts_sublist_sorted[self.container_counter] = \
parts_sublist[self.container_counter][self.sublist_index_order]
# old_index_order = np.argsort(index_order)
# Initialize branch and bound matrix, bbm, with level 1 node
# Row will be [level a0_1 a0_2 ... a0_n rem LP IP]
# bbm will consist of all nodes that may still be explored
self.bbm = np.zeros((1, bandwidth + 4))
self.bbm[0, 0] = 1 # level
# Entries 1 through num_parts will remain at 0 for the first node because no parts have been nested
self.bbm[0, bandwidth + 1] = self.nestable_length # Calculate rem, remaining nestable length
self.bbm[0, bandwidth + 2] = self.bbm[0, bandwidth + 1] * self.values_sublist_sorted[0] # Calculate
# value of LP, linear programming maximum value
# bbm[0, num_parts + 3] = 0 # Value of IP, will remain at 0 since no parts are nested
# Initialize lp_best and ip_best
self.lp_best = self.bbm[0, bandwidth + 2]
self.ip_best = 0
# Initialize ip_best_row to keep track of best node (the one with highest IP value)
self.ip_best_row = self.bbm[0]
# Initialize lp_best_index at 0 since there is only one node
self.lp_best_index = 0
# Initialize loop_count
self.loop_count = 0
# Begin loop to start branching nodes, and allow for rounding error
while self.lp_best > self.ip_best - 0.00000001:
# Check if calculation has been canceled.
if self.calculation_was_canceled == 1:
# Zero out all outputs and exit function
self.final_patterns = []
self.final_allocations = 0
return 0
# return self.required_lengths, self.allocation, self.patterns
# # return self.final_patterns, self.final_allocations
# Extract the row to be explored
self.row = self.bbm[self.lp_best_index, :]
# Extract the level of the row to be explored
self.level = int(self.row[0])
# Extract the remaining length for the row to be explored
self.rem = self.row[bandwidth + 1]
# Extract length of part being considered at current level (level 1 corresponds to part 1 and so on)
self.p_length = self.nested_lengths_sublist_sorted[self.level - 1]
# Check how many of the part can be nested on remaining length rem
self.num = math.floor(self.rem / self.p_length)
# Reduce num if there are not enough parts available in the job to add to the nest
self.part_max = int(math.floor(self.part_quantities_sublist_sorted[self.level - 1].item()))
if self.part_max < self.num:
self.num = self.part_max
# Allow node to be explored by default
# Set a variable to 1 to allow the node to be branched into sub-nodes
self.branch_node = 1
# Limit the number of parts that can be used in a pattern
if self.level > max_parts_per_nest:
# Count number of different parts in the pattern so far
self.parts_in_pattern = 0
for i in range(self.level):
if self.row[i + 1] != 0:
self.parts_in_pattern += 1
# Check when only one more part can be added to the pattern
if self.parts_in_pattern == max_parts_per_nest - 1:
# Copy the node 1 time, and iterate a0_level to num since no other parts will be nested later
self.row_copy = self.row.copy() # copy row
self.row_copy[self.level] = self.num # iterate a0_level
self.row_copy[bandwidth + 1] = self.rem - self.num * self.p_length # subtract nested parts from rem
self.row_copy[bandwidth + 3] = np.dot([self.row_copy[1:bandwidth + 1]],
self.pi_sublist_sorted) # calculate IP for new node
# Check if current pattern fulfills subsequent uff parts without creating more uff parts
# upstream. Give bonus IP when this occurs.
if mode == 1:
bonus = self.check_for_bonus_condition()
if bonus == 1:
self.row_copy[bandwidth + 3] += 0.08
# Add slight incentive to reduce number of different parts in each pattern (prevents unnecessary
# mixing
num_parts_in_nest = 0
for value in self.row_copy[1:(bandwidth + 1)]:
if value:
num_parts_in_nest += 1
if num_parts_in_nest > 0:
self.row_copy[bandwidth + 3] += 1 - num_parts_in_nest / (num_parts_in_nest - 0.001)
if self.row_copy[bandwidth + 3] > self.ip_best:
self.ip_best = self.row_copy[bandwidth + 3]
self.ip_best_row = self.row_copy
# If no parts were added to the nest, keep branching.
if self.num == 0:
self.branch_node = 1
else:
self.branch_node = 0
# Remove the explored node from bbm
self.bbm = np.delete(self.bbm, self.lp_best_index, 0)
# Iterate original row to next level to allow for other parts to be added as final part instead
if self.level != bandwidth:
self.row[0] = self.level + 1
self.bbm = np.append(self.bbm, [self.row], axis=0) # add new node to bbm
elif self.parts_in_pattern == max_parts_per_nest:
self.bbm = np.delete(self.bbm, self.lp_best_index, 0)
self.branch_node = 0
if self.branch_node == 1:
# Copy the node (num + 1) times to explore it, and iterate the qty for the current part from 0 to num
selected_range = range(self.num + 1)
# Only consider the option of using all remaining parts if considering the current unfulfilled part
if mode == 1 and self.level == 1:
selected_range = range(self.num, self.num + 1)
for i in selected_range:
self.row_copy = self.row.copy() # copy row
self.row_copy[self.level] = i # iterate qty for the current part
self.row_copy[bandwidth + 1] = self.rem - i * self.p_length # subtract nested parts from rem
self.row_copy[bandwidth + 3] = np.dot([self.row_copy[1:bandwidth + 1]],
self.pi_sublist_sorted) # calculate IP for new node
# Check if current pattern fulfills subsequent uff parts without creating more uff parts
# upstream. Give bonus IP when this occurs.
if mode == 1:
bonus = self.check_for_bonus_condition()
if bonus == 1:
self.row_copy[bandwidth + 3] += 0.08
# Add slight incentive to reduce number of different parts in each pattern (prevents unnecessary
# mixing
num_parts_in_nest = 0
for value in self.row_copy[1:(bandwidth + 1)]:
if value:
num_parts_in_nest += 1
if num_parts_in_nest > 0:
self.row_copy[bandwidth + 3] += 1 - num_parts_in_nest / (num_parts_in_nest - 0.001)
if self.row_copy[bandwidth + 3] > self.ip_best:
self.ip_best = self.row_copy[bandwidth + 3]
self.ip_best_row = self.row_copy
if self.level < bandwidth:
self.row_copy[bandwidth + 2] = self.row_copy[bandwidth + 3] + self.row_copy[
bandwidth + 1] * self.values_sublist_sorted[
self.level] # calculate LP for new node
if self.row_copy[bandwidth + 2] > self.ip_best:
self.row_copy[0] = self.level + 1 # increment level of copy
self.bbm = np.append(self.bbm, [self.row_copy], axis=0) # add new node to bbm
# Remove the explored node from bbm
self.bbm = np.delete(self.bbm, self.lp_best_index, 0)
# Every 100 iterations, check for any nodes with an LP less than ip_best and remove them
if self.loop_count / 100 == round(self.loop_count / 100):
for i in (range(len(self.bbm[:, bandwidth + 2])))[::-1]:
if self.bbm[i, bandwidth + 2] < self.ip_best:
self.bbm = np.delete(self.bbm, i, 0)
# Check to make sure bbm is not empty
if np.size(self.bbm) > 0:
# TODO Find a faster way to choose next node without cycling through all of bbm
# Decide which node to explore next by searching for the node in bbm with highest LP
self.lp_best_index = np.argmax(self.bbm[:, bandwidth + 2])
self.lp_best = self.bbm[self.lp_best_index, bandwidth + 2]
else:
break
if self.loop_count == 10000:
break
# Keep track of how many times the loop is executed
self.loop_count += 1
def column_gen(self, bandwidth, max_parts_per_nest, part_quantities, parts_sublist, limit_iterations):
# Initialize patterns matrix (single part nesting patterns)
self.patterns = np.zeros((self.num_parts, self.num_parts))
for i in range(self.num_parts):
self.patterns[i, i] = min(self.window.part_quantities[i],
math.floor(self.nestable_length / self.nested_lengths[i]))
# start_time_cg = time.time() + 1000
# start_time_cg = time.time()
# Reset counters
self.container_counter = 0
self.iteration_count = 0
# Initialize a tracker to check for periodic cycling in ip_best
# TODO make this less arbitrary, should be based on bandwidth too
ip_best_history = np.array(range(self.num_parts * 20))
# TODO find a way to check if minimum is global or local, try different initial conditions?
# Execute main program loop until optimum solution is reached
# while time.time() - start_time_cg < time_limit:
while self.iteration_count < self.max_iterations:
if self.calculation_was_canceled == 1:
# Zero out all outputs and exit function
self.final_patterns = []
self.final_allocations = 0
return self.required_lengths, self.allocation, self.patterns
self.branch_bound(bandwidth, max_parts_per_nest, part_quantities, parts_sublist, 0)
if self.iteration_count % 10 == 0:
print(f"Number of passes during branch_bound: {self.loop_count}")
# Check if calculation has been canceled.
if self.calculation_was_canceled == 1:
# Zero out all outputs and exit function
self.final_patterns = []
self.final_allocations = 0
return self.required_lengths, self.allocation, self.patterns
if self.ip_best >= 1:
# Extract the best pattern from ip_best_row
self.best_pattern_sorted_sublist = np.transpose([self.ip_best_row[1:(bandwidth + 1)]])
# Initialize best_pattern_sorted
self.best_pattern_sorted = np.zeros((self.num_parts, 1))
# Add pattern values from the sublist to the main list
for i in range(bandwidth):
# Find index in main list corresponding to ith index of parts_sublist_sorted
self.corr_index = \
np.where(self.index_order == self.parts_sublist_sorted[self.container_counter][i])[0].item()
self.best_pattern_sorted[self.corr_index] = self.best_pattern_sorted_sublist[i]
# Reorder best_pattern vector
self.old_index_order = np.argsort(self.index_order)
self.best_pattern = self.best_pattern_sorted[self.old_index_order]
# Determine which pattern to replace in patterns
# Solve for p_bar_j (proportion of each existing pattern that 1 instance of best_pattern can replace)
self.p_bar_j = np.dot(self.patterns_inv, self.best_pattern)
# Initialize theta_limits
self.theta_limits = np.zeros((self.num_parts, 1))
# Calculate required_lengths before adding new pattern to patterns
self.required_lengths = np.sum(np.dot(self.patterns_inv, part_quantities))
# Find limiting pattern when replacing current nests with the max number of instances of best_pattern
for i in range(self.num_parts):
# Solve for the reciprocal of the limits on Theta for each pattern
self.theta_limits[i] = self.p_bar_j[i] / (self.allocation[i] + 0.00000001)
# Replace the pattern with the largest value of theta_limits
self.index_to_replace = np.argmax(self.theta_limits)
self.patterns[:, self.index_to_replace] = self.best_pattern.transpose()[0]
# Fixes issue with allocation not matching new version of pattern TODO figure out why
self.patterns_inv = np.linalg.inv(self.patterns)
self.allocation = np.dot(self.patterns_inv, part_quantities)
# Update ip_best_history
ip_best_history = np.append(ip_best_history[1:], self.ip_best)
self.iteration_count += 1
# If repetitions are found in ip_best_history, then exit the column generation function with results.
if self.find_cycling(ip_best_history):
# print(time.time() - start_time_cg)
print(f"cycling found after {self.iteration_count} iterations")
# decrement max_iterations to make cycling less likely
if limit_iterations == 1:
if self.max_iterations == 1000000: # TODO recode since this condition is only met on first loop
self.max_iterations = self.iteration_count
else:
self.adjust_max_iterations(0.95)
return self.required_lengths, self.allocation, self.patterns
# Reset container_counter when it reaches num_parts - bandwidth + 1
self.container_counter += 1
if self.container_counter == self.num_parts - bandwidth + 1:
self.container_counter = 0
print(f"reached maximum iterations of {self.max_iterations}")
# increment max_iterations to make cycling more likely
if limit_iterations == 1:
self.adjust_max_iterations(1.01)
return self.required_lengths, self.allocation, self.patterns
# def decrease_max_iterations(self):
# new_max_iterations = math.floor(self.max_iterations * 0.95)
# if self.max_iterations == new_max_iterations:
# self.max_iterations -= 1
# else:
# self.max_iterations = new_max_iterations
def run(self):
nesting_start_time = time.time()
[final_patterns, final_allocations] = self.length_nest_pro_calculate()
results = [final_patterns, final_allocations]
print(f"nesting time was {time.time() - nesting_start_time}")
# noinspection PyUnresolvedReferences
self.results_signal.emit(results)
# TODO add algorithm/option that tends to pick nests that use up a single part faster (reduces the dependency of the
# patterns on the part quantities)
# TODO add algorithm/option to reduce the number of containers needed (completely finish first 3 parts before
# starting on the next parts. Never cut the nth part if the (num_parts-3)th part is not completed. Allow user
# to adjust 3 to other values.
# TODO add functionality to calculate optimum stock length (by iterating with different stock lengths?)
# TODO allow user to select multiple stock lengths and quantities/priorities
# TODO remove timers
# TODO add file about for version info and help
# Create function to nest parts
def length_nest_pro_calculate(self):
##########################
# Pre-processing section #
##########################
# Start timer
# nesting_start_time = time.time()
self.max_iterations = 1000000
# Check how many different parts are needed (num_parts)
self.num_parts = len(self.window.part_lengths)
# Set precision for printing
np.set_printoptions(precision=3)
# Remove any parts where the part quantity is 0 (remove entries from self.window.part_names,
# self.window.part_lengths, and self.window.part_quantities)
for i in range(len(self.window.part_quantities))[::-1]:
if self.window.part_quantities[i] == 0:
self.window.part_quantities = np.delete(self.window.part_quantities, i, 0)
self.window.part_lengths = np.delete(self.window.part_lengths, i, 0)
self.window.part_names = np.delete(self.window.part_names, i, 0)
initial_part_quantities = self.window.part_quantities.copy()
initial_part_lengths = self.window.part_lengths.copy()
initial_part_names = self.window.part_names.copy()
# Update number of parts after removing parts with qty 0
self.num_parts = len(self.window.part_lengths)
# Make sure max_containers is not higher than num_parts, and make sure it was not a string.
if self.window.max_containers > self.num_parts or self.window.max_containers == -2:
self.window.max_containers = self.num_parts
# Make sure max_parts_per_nest is not greater than max_containers since that wouldn't make sense.
# Also make sure it was not entered as a string
if self.window.max_parts_per_nest > self.window.max_containers or self.window.max_parts_per_nest == -2:
self.window.max_parts_per_nest = self.window.max_containers
# TODO Combine parts with same length? Must still consider quantities...
# TODO Find all optimum nodes, not just one
# Solve for nestable length with extra spacing adjustment since blank includes spacing
initial_nestable_length = \
self.window.stock_length \
- self.window.left_waste \
- self.window.right_waste \
+ self.window.spacing
self.nestable_length = initial_nestable_length
# Construct length vector by adding part spacing
self.nested_lengths = np.zeros((self.num_parts, 1))
for i in range(self.num_parts):
self.nested_lengths[i, 0] = self.window.part_lengths[i, 0] + self.window.spacing
initial_nested_lengths = self.nested_lengths.copy()
# Initialize patterns matrix (Step #1) (single part nesting patterns)
self.patterns = np.zeros((self.num_parts, self.num_parts))
for i in range(self.num_parts):
self.patterns[i, i] = math.floor(self.nestable_length / self.nested_lengths[i])
# Check if each part can be nested on the available length
if self.patterns[i, i] == 0 or self.window.stock_length < self.window.left_waste + self.window.right_waste:
# Zero out all outputs and exit function
self.final_patterns = []
self.final_allocations = 0
self.window.error = 1 # error code 1 signifies that a part is too long
return self.final_patterns, self.final_allocations
# Find required number of lengths if everything nests ideally (only possible if parts nest perfectly on nestable
# length)
self.ideal_num = (np.dot(np.transpose(self.nested_lengths),
self.window.part_quantities) / self.nestable_length).item()
print("\nIdeally, the job would only require " + str(round(self.ideal_num, 2)) + " lengths. (zero scrap)\n")
# Find required number of lengths in worst case scenario (single part nests only)
self.patterns_inv = np.linalg.inv(self.patterns)
self.patterns_trans_inv = np.transpose(self.patterns_inv)
self.ones_vector = np.ones((self.num_parts, 1))
# pi is a measure of how much of a stock length is used to cut a given part (if the 3rd term is 0.25, that
# would indicate that the third part uses 1/4 of a stock length when considering the entire nest with all parts)
self.pi = np.dot(self.patterns_trans_inv, self.ones_vector)
self.worst_case = np.dot(np.transpose(self.window.part_quantities), self.pi)
self.required_lengths = self.worst_case.copy()
print("If only single part nests are used, the job would require a maximum of " + str(
round(self.worst_case.item(), 2))
+ " lengths.\n")
# Initialize parts_sublist (will restrain column generation to only consider parts spanning range of
# max_containers)
self.parts_sublist = np.zeros((self.num_parts - self.window.max_containers + 1, self.window.max_containers))
self.parts_sublist_sorted = np.zeros((1, self.num_parts))
self.container_counter = 0
self.remaining_iterations = -1
# TODO remove this feature if it is not desired (speeds up convergence, but results are less consistent)
# # Initialize the part ordering if it doesn't exist yet
# try:
# print(self.window.current_sequence)
# except AttributeError:
# self.window.current_sequence = np.array(range(self.num_parts))
# Initialize the part ordering
self.window.current_sequence = np.arange(self.num_parts)
for sub_i in range(self.num_parts - self.window.max_containers + 1):
self.parts_sublist[sub_i] = np.array(range(sub_i, sub_i + self.window.max_containers))
# self.parts_sublist[sub_i] = self.window.current_sequence[sub_i:(sub_i + self.window.max_containers)]
# Iterate the part sequence to find the part ordering for which scrap is minimized
self.part_sequence_is_optimum = 0
self.current_part_index = 0
self.window.part_quantities = initial_part_quantities[self.window.current_sequence].copy()
self.window.part_lengths = initial_part_lengths[self.window.current_sequence].copy()
self.window.part_names = initial_part_names[self.window.current_sequence].copy()
self.parts_sublist_sorted = np.zeros((self.num_parts - self.window.max_containers + 1,
self.window.max_containers))
best_sequence = self.window.current_sequence.copy()
# best_required_lengths = self.worst_case.copy()
# part_sequence_is_optimum = 0
# ij_best = [-1, -1]
# # Don't enter while loop if max_containers is 1
# if self.window.max_containers == 1:
# part_sequence_is_optimum = 1
# sequencing_loop_counter = 0
# sequencing_loop_counter_best = 0
# x = []
# y = []
# # Loop to find best sequence of parts
# while part_sequence_is_optimum == 0:
# for i in range(self.num_parts):
# if sequencing_loop_counter - sequencing_loop_counter_best >= 6:
# part_sequence_is_optimum = 1
# break
# if part_sequence_is_optimum == 1:
# break
# starting_sequence = self.window.current_sequence.copy()
# for j in range(self.num_parts):
# if part_sequence_is_optimum == 1:
# break
# index_to_move = np.where(starting_sequence == i)[0]
#
# # Move current part to jth position (swap)
# term_a = starting_sequence[index_to_move].copy()
# term_b = starting_sequence[j].copy()
# self.window.current_sequence = starting_sequence.copy()
# self.window.current_sequence[index_to_move] = term_b.copy()
# self.window.current_sequence[j] = term_a.copy()
#
# self.window.part_quantities = initial_part_quantities[self.window.current_sequence].copy()
# self.window.part_lengths = initial_part_lengths[self.window.current_sequence].copy()
# self.window.part_names = initial_part_names[self.window.current_sequence].copy()
#
# self.nested_lengths = initial_nested_lengths[self.window.current_sequence].copy()
#
# # Reinitialize nestable length
# self.nestable_length = initial_nestable_length
#
# # Reinitialize patterns matrix (single part nesting patterns)
# self.patterns = np.zeros((self.num_parts, self.num_parts))
# for k in range(self.num_parts):
# self.patterns[k, k] = math.floor(self.nestable_length / self.nested_lengths[k])
#
# # Run the column generation algorithm to find the best set of nest patterns for the job quantities
# [self.required_lengths, self.allocation, self.patterns] = \
# self.column_gen(self.window.max_containers, self.window.part_quantities,
# self.parts_sublist, 0)
#
# # Check if i and j match from the last time a best sequence was found
# if [i, j] == ij_best:
# part_sequence_is_optimum = 1
#
# if self.required_lengths < best_required_lengths - 0.00001:
# print("new best")
# print(self.required_lengths)
# best_required_lengths = self.required_lengths.copy()
# ij_best = [i, j]
# sequencing_loop_counter_best = sequencing_loop_counter
# best_sequence = self.window.current_sequence.copy()
#
# x = np.append(x, time.time())
# y = np.append(y, self.required_lengths)
#
# # Is this needed? It seems to help reduce scrap in some cases.
# elif self.required_lengths == best_required_lengths:
# print("tied best")
# print(self.required_lengths)
# best_required_lengths = self.required_lengths.copy()
# best_sequence = self.window.current_sequence.copy()
#
# self.window.current_sequence = best_sequence.copy()
#
# sequencing_loop_counter += 1
#
# x = np.append(x, time.time())
# y = np.append(y, self.required_lengths)
#
# plt.plot(x, y)
# plt.savefig("mygraph.png")
# Run through one more time using best sequence to get best patterns
self.window.part_quantities = initial_part_quantities[best_sequence].copy()
self.window.part_lengths = initial_part_lengths[best_sequence].copy()
self.window.part_names = initial_part_names[best_sequence].copy()
self.nested_lengths = initial_nested_lengths[best_sequence].copy()
# Reinitialize nestable length
self.nestable_length = initial_nestable_length
# Reinitialize patterns matrix (single part nesting patterns)
self.patterns = np.zeros((self.num_parts, self.num_parts))
for k in range(self.num_parts):
self.patterns[k, k] = math.floor(self.nestable_length / self.nested_lengths[k])
# Run the column generation algorithm to find the best set of nest patterns for the job quantities
# Decrease max parts per nest until max containers is fulfilled.
for i in range(1, 1 + self.window.max_parts_per_nest)[::-1]:
temp_max_parts_per_nest = i
[self.required_lengths, self.allocation, self.patterns] = \
self.column_gen(self.num_parts, temp_max_parts_per_nest, self.window.part_quantities,
np.array([np.arange(self.num_parts)]), 0)
# Check if calculation has been canceled.
if self.calculation_was_canceled == 1:
# Zero out all outputs and exit function
self.final_patterns = []
self.final_allocations = 0
return self.final_patterns, self.final_allocations
# Round all of the allocations down, but allow for rounding error
self.int_allocation = self.allocation.copy()
for ii in range(len(self.allocation))[::-1]:
self.int_allocation[ii] = math.floor(self.allocation[ii] + 0.0000001)
# Remove unused patterns
if self.int_allocation[ii] == 0:
self.int_allocation = np.delete(self.int_allocation, ii, 0)
self.allocation = np.delete(self.allocation, ii, 0)
self.patterns = np.delete(self.patterns.T, ii, 0).T
# TODO move this to a better place
cs = ColumnSorter(self.num_parts, len(self.allocation), self)
if self.patterns.any():
[new_column_order, required_containers] = cs.optimize_sequence(self.patterns, 100, 0)
# TODO remove second condition later?
if required_containers <= self.window.max_containers or temp_max_parts_per_nest == 2:
self.patterns = self.patterns.T[new_column_order].T
self.int_allocation = self.int_allocation[new_column_order]
chosen_required_lengths = self.required_lengths.copy()
chosen_int_allocation = self.int_allocation.copy()
chosen_patterns = self.patterns.copy()
print(f"Acceptable solution found with 'Max parts per nest' constrained to {i}")
print(f"Requires about {self.required_lengths} lengths")
break
else:
chosen_required_lengths = self.required_lengths.copy()
chosen_int_allocation = self.int_allocation.copy()
chosen_patterns = self.patterns.copy()
print(f"No integer solution was found with 'Max parts per nest' constrained to {i}")
break
# TODO check for best solution before filling remaining parts
self.required_lengths = chosen_required_lengths.copy()
self.int_allocation = chosen_int_allocation.copy()
self.patterns = chosen_patterns.copy()
# Create active parts matrix, and find start_pattern and end_pattern for each part inside pre_process function
self.active_parts = self.patterns.copy()
self.active_parts = cs.pre_process(self.active_parts)
# Sort parts (rows) by end_pattern (earliest to latest)
sorted_by_start = np.argsort(cs.start_pattern)
cs.end_pattern_sorted = cs.end_pattern[sorted_by_start]
cs.end_pattern_sorted = cs.end_pattern_sorted.astype(float)
for item_index, item in enumerate(cs.end_pattern_sorted):
cs.end_pattern_sorted[item_index] = item + item_index * 0.0000000001
sorted_by_end = np.argsort(cs.end_pattern_sorted)
self.active_parts = self.active_parts[sorted_by_start][sorted_by_end]
self.patterns = self.patterns[sorted_by_start][sorted_by_end]
self.window.part_quantities = self.window.part_quantities[sorted_by_start][sorted_by_end]
self.window.part_names = self.window.part_names[sorted_by_start][sorted_by_end]
self.window.part_lengths = self.window.part_lengths[sorted_by_start][sorted_by_end]
self.nested_lengths = self.nested_lengths[sorted_by_start][sorted_by_end]
cs.start_pattern = cs.start_pattern[sorted_by_start][sorted_by_end]
cs.end_pattern = cs.end_pattern[sorted_by_start][sorted_by_end]
self.frozen_pi = self.pi[sorted_by_start][sorted_by_end]
containers = cs.count_containers(self.active_parts)
containers_equals_max = containers == self.window.max_containers
# Subtract quantities of fully allocated sticks to find remaining_part_quantities
self.nested = np.dot(self.patterns, self.int_allocation)
self.remaining_part_quantities = self.window.part_quantities.copy()
self.remaining_part_quantities = self.remaining_part_quantities - self.nested
# Initialize with 2 dummy terms
self.additional_patterns = np.zeros((self.num_parts, 2))
self.additional_allocations = np.zeros((1, 2))
# TODO analyze how well this bug fix works and improve
# maybe try simple allocation without nesting
# Fix cs.start_pattern and cs.end_pattern if they contain terms that are out of bounds
acceptable_values = np.arange(len(self.int_allocation))
skip_to_end = 0
for num_index, num in enumerate(cs.start_pattern):
if num not in acceptable_values:
skip_to_end = 1
# Initialize with dummy term
bridge_position_tracker = np.array([[0]])
# Loop through parts (top to bottom)
for self.part_index, active_parts_row in enumerate(self.active_parts):
# If part is unfulfilled, generate a bridge pattern to fulfill it without breaking max_containers req
# Use the nest with the lowest possible scrap rate (greedy algorithm),
# but add preference for fulfilling other unfulfilled parts
# Stop early if remaining parts can be nested without breaking max_containers requirement
if self.part_index >= (self.num_parts - self.window.max_containers):
break
if skip_to_end == 1:
break
if self.remaining_part_quantities[self.part_index]:
# Find index of next unfulfilled part after current part
next_uff_part = self.num_parts - 1
for ii in range(self.part_index + 1, self.num_parts):
if self.remaining_part_quantities[ii]:
next_uff_part = ii
break
# Define left and right bounds for current parts
left_bound = int(cs.start_pattern[self.part_index])
right_bound = int(cs.end_pattern[next_uff_part])
# Adjust right bound to avoid breaking max_containers requirement
for j in range(left_bound, right_bound):
# Using j as the left pattern under the below conditions would break the max_containers requirement
if containers_equals_max[j] and active_parts_row[j] == 0:
right_bound = j
break
if left_bound == right_bound:
left_is_same_as_right = 1
else:
left_is_same_as_right = 0
# Loop through pairs of patterns within bounds to define parts that may be included in bridge patterns
bridge_sublist = np.zeros((right_bound + left_is_same_as_right - left_bound, self.num_parts))
for j_index, j in enumerate(range(left_bound, right_bound + left_is_same_as_right)):
left_active = self.active_parts[:, j]
right_active = self.active_parts[:, j + 1 - left_is_same_as_right]
# Current part should always be used in bridge pattern
bridge_sublist[j_index][self.part_index] = 1
# Other parts are allowed in bridge pattern if they are adjacent to an active part
# Other parts must be below current part
for ii in range(self.part_index + 1, self.num_parts):
if left_active[ii] or right_active[ii]:
bridge_sublist[j_index][ii] = 1
# bridge_sublist = np.unique(bridge_sublist, axis=0)
# print(bridge_sublist) # TODO add similar functionality later
self.unavailable_parts = [[-1]]
try_expanding_right_bound = 0
num_extra_parts = 0
while self.remaining_part_quantities[self.part_index]: # Finds additional bridge pattern if needed
if try_expanding_right_bound == 1 and right_bound < np.shape(self.active_parts)[1] - 2:
right_bound += 1
bridge_sublist = np.append(bridge_sublist, np.zeros((1, self.num_parts)), axis=0)
# Loop through pairs of adjacent patterns within bounds to define parts that may be included in
# each bridge patterns
left_active = self.active_parts[:, right_bound]
right_active = self.active_parts[:, right_bound + 1]
# Current part should always be used in bridge pattern
bridge_sublist[-1][self.part_index] = 1
# Other parts are allowed in bridge pattern if they are adjacent to an active part
# Other parts must be below current part
for ii in range(self.part_index + 1, self.num_parts):
if left_active[ii] or right_active[ii]:
bridge_sublist[j_index][ii] = 1
best_bridge_pattern = []
best_bridge_ip = 0
for sublist_index, sublist in enumerate(bridge_sublist):
parts_sublist = np.array([])
for bit_index, bit in enumerate(sublist):
if bit == 1:
parts_sublist = np.append(parts_sublist, bit_index)
parts_sublist = np.array([parts_sublist.astype(int)])
self.container_counter = 0
# Create new sublist that helps branch bound function decide which nests are best
self.bonus_sublist = parts_sublist.copy()
for item_index, item in enumerate(self.bonus_sublist[0]):
if item == self.part_index or item == next_uff_part:
self.bonus_sublist[0][item_index] = 1
elif item > next_uff_part:
self.bonus_sublist[0][item_index] = 0
else:
self.bonus_sublist[0][item_index] = -1
self.branch_bound(len(parts_sublist[0]), self.window.max_parts_per_nest + num_extra_parts,
self.window.part_quantities.copy(), parts_sublist, 1,
np.where(parts_sublist[0] == self.part_index)[0].item())
# Check if calculation has been canceled.
if self.calculation_was_canceled == 1:
# Zero out all outputs and exit function
self.final_patterns = []
self.final_allocations = 0
return self.final_patterns, self.final_allocations
# Extract the best pattern from ip_best_row
self.best_pattern_sorted_sublist = \
np.transpose([self.ip_best_row[1:(len(parts_sublist[0]) + 1)]])
# Initialize best_pattern_sorted
self.best_pattern_sorted = np.zeros((self.num_parts, 1))
# Add pattern values from the sublist to the main list
for iii in range(len(parts_sublist[0])):
# Find index in main list corresponding to ith index of parts_sublist_sorted
self.corr_index = \
np.where(self.index_order == self.parts_sublist_sorted[self.container_counter][iii])[
0].item()
self.best_pattern_sorted[self.corr_index] = self.best_pattern_sorted_sublist[iii]
# Reorder best_pattern vector
self.old_index_order = np.argsort(self.index_order)
self.best_pattern = self.best_pattern_sorted[self.old_index_order]
if self.ip_best > best_bridge_ip:
best_bridge_ip = self.ip_best
best_bridge_pattern = self.best_pattern.copy()
insertion_position = cs.start_pattern[self.part_index] + sublist_index + 1
# Check if any downstream unfulfilled parts could have fit on current bridge pattern
remaining_length = self.window.stock_length - self.window.left_waste - self.window.right_waste
remaining_length -= np.dot(best_bridge_pattern.T, self.nested_lengths)
downstream_parts = np.arange(self.part_index + 1, self.num_parts)
find_new_best = 0
for p_index, part in enumerate(downstream_parts):
if not best_bridge_pattern[part][0]: # If the part is not in the bridge pattern
if self.nested_lengths[p_index] < remaining_length: # and if there is room to nest it
if right_bound < np.shape(self.active_parts)[1] - 2: # and the right bound can be
# extended
try_expanding_right_bound = 1
find_new_best = 1
num_parts_in_nest = 0
for p in best_bridge_pattern:
if p[0]:
num_parts_in_nest += 1
if num_parts_in_nest == self.window.max_parts_per_nest + num_extra_parts:
num_extra_parts += 1
break
if find_new_best == 1:
continue
else:
num_extra_parts = 0
# Save copies of changing variables in case they have to change back
backup_vars = [bridge_position_tracker.copy(),
self.additional_patterns.copy(),
self.additional_allocations.copy(),
self.remaining_part_quantities.copy(),
self.int_allocation.copy()
]
bridge_position_tracker = np.append(bridge_position_tracker, insertion_position)
self.additional_patterns = np.append(self.additional_patterns.T, best_bridge_pattern.T, axis=0).T
self.additional_allocations = np.append(self.additional_allocations.T, [[1]], axis=0).T
self.remaining_part_quantities -= best_bridge_pattern
do_not_adjust_allocations = 0
# Loop to decide which patterns to borrow from
while np.any(self.remaining_part_quantities < 0):
for pattern_qty_index, pattern_qty in enumerate(best_bridge_pattern):
if pattern_qty:
if pattern_qty_index <= self.part_index: # Catches part_index part
continue
# Borrow from the rightmost possible pattern if there are negative parts remaining
if self.remaining_part_quantities[pattern_qty_index] < 0:
rightmost_pattern_index = cs.end_pattern[pattern_qty_index]
# Make sure there are patterns available to pull from before borrowing parts
if self.int_allocation[rightmost_pattern_index] == 0:
# Keep trying patterns to the left until there are allocations for that pattern
# and it uses the part with negative qty remaining
while self.int_allocation[rightmost_pattern_index] == 0 or \
self.patterns[pattern_qty_index][rightmost_pattern_index] == 0:
rightmost_pattern_index = rightmost_pattern_index - 1
# TODO come up with a better solution here
if rightmost_pattern_index == -1:
# Something must be wrong in code if this happens, but skipping to end
# should avoid a crash
skip_to_end = 1
else:
for qty_index, qty in \
enumerate(self.patterns.T[rightmost_pattern_index]):
if qty:
highest_in_pattern = qty_index
break
# If pattern contains any parts with index <= self.part_index (higher